Computer Oral History Collection, , 1977

Transcription

1 Computer Oral History Collection, , 1977 Interviewee: Howard Aiken ( ) Interviewers: Henry Tropp and I.B. Cohen Date: February 26-27, 1973 Repository: Archives Center, National Museum of American History This is the 26th of February 1973, and we're holding a discussion with Dr. Howard Aiken in his home in Ft. Lauderdale. The other participants are Professor I.B. Cohen and Henry Tropp. Would you like to begin, Professor Cohen? Sure. Well, I've never done a taped interview like this, exactly. Neither have I. I think, therefore, that I ought to say that I would like to put my terms of participation on record. And they are, that in the first place, that quite clearly, whatever we record here is the literary property of H.H. Aiken, and I think it has to be understood that if he says when we are finished, "You know, I was indiscrete", or something, "and I think we better not let anyone ever hear that until I am out of this world or out of this world for a time", or whatever he wants to do. As a historian, I've got to go on record that this is said. Well, I'm glad you went on record, but this is standard procedure for my operation: that this is a privileged piece of information until Professor Aiken decides what use, if any, can be made of it. And I'll add at this point that unless I think that it is a mistake to add further remarks on watches, (Laughter) I think that's a comment almost worth recording.

2 Computer Oral History Collection, , Well, let me begin by two rather general questions. I've made a few more specific ones here. This is something I think I really already know, but I think that I would like to get it authenticated. The story is usually told, which as I recall, I've heard you tell, is that you really became concerned with problems of mechanical computation during the thesis you wrote with Leon Chaffee. That's correct, yes. That is right, isn't it? Yes. I had done a thesis on space charge, and this is a field where one runs into cylindrical coordinates, or in parallel cases, into ordinary differential equations in nonlinear terms, of course. Actually, the object of the thesis almost became solving nonlinear equations. Not completely, but there was some of that in it. The only methods available were methods and they are extremely time consuming. And it became apparent to me at once that this could be mechanized and programmed and that an individual didn't have to do this. And it was at that time that it became clear that in the words that I used then, there were different kinds of computing procedures computing by columns, in which one listed all the values of the independent variable, and then made all the first steps and then all the second steps and so on. And the punched card machine was really ideal for this. One didn't have to do a thing. when the data had the n plus the first row of the table you had to compute across this and this and then this., which no other computing device at the time could do. And this was so elementary and it seemed so simple, that you wonder why I take the time to discuss it, but it was a completely personal observation of the first importance to me at that time. Because in the analog machinery, there was nothing about this. That last point determined the name of the game when we got controls. Yes. But even at that early stage, you were aware then, I take it, of the punched cards.

3 Computer Oral History Collection, , Oh, yes, that's right. Why? Why was I aware of them? Yes. I wondered how many people. Well now, Leon Chaffee, for example, would he have been aware of them? No. I wouldn't have thought so. Well, I became aware of it because having decided that something should be done of this. I found in the library these two books: Napier's "Exhibition" My favorite book. That thing, and even more important in some ways, "The Catalogue of the South Kensington Museum." And those two things provided me with an enormous amount of background information. But these I didn't find at once, and this led to a rather peculiar amusing incident. The faculty had rather limited enthusiasm about what I wanted to do, if not almost downright antagonism, and we had an assistant in the Department of Physics. If I remember right, his name was Lancsos, Can you remember? Cornelius Lancsos? He's at the Institute for Advanced Study at Dublin now. Was he at Harvard then?

4 Computer Oral History Collection, , Oh, no, no. This was in the Laboratory. Lanza. Lanza. That was the name. Yes. There were three of them. That's right. Cornelius was one. Yes, there were three brothers, and they were mechanics. That's right. Well, anyway, after the proposals for the Napier Mark I had been set forth and people were talking about it, and the faculty had begun to make it rather clear then, that they had no interest, I guess it was Saunders who mentioned it to Lanza, who added his viewpoint. And he couldn't see why in the world I wanted to do anything like this in the Physics Laboratory, because we already had such a machine and nobody ever used it. I heard about this, and demanded to know where we had this machine and it was in the attic of the old research laboratories. And sure enough, we had two of Babbage's wheels. Those were the wheels that I had later mounted and put in the body of the computer. But this was the first time I had ever heard of Babbage, and then, from there, I found Babbage in this thing: "Passages of Life in the ", and from that, I got it out of these, and there's my education in computers right there, this is the whole thing. Everything I took out of a book. That's practically all there was. There really wasn't anything else.

5 Computer Oral History Collection, , Well, that's extremely interesting, because I've often wondered about that part of it. In fact, I was going to ask you another question, which I don't need to anymore, whether you had ever by chance on a vacation, gone to the South Kensington Museum. I went to the South Kensington Museum after the war, while the place was in still a shambles. Well, I went to England and I always stayed with the Comries, and Comrie made arrangements for the Museum to be opened up so that I could get up and see the analytical engine and it was the greatest disappointment of my life. But that's the way I got to see it. Did you know Comrie? It was one of the great regrets of my life that I didn't. I've read his papers and what a wonderful man he was. Oh, he was fantastic. He was a crusty so-and-so, you know. When he published a mathematics paper, he always put a number of round-off errors in it, and then recorded these errors and wrote a letter to himself which went through the mails, and then deposited that letter with the Secretary of the Royal Society. So that when subsequently, somebody copied his table, he could have this envelope opened and show them the round-off error. (Laughter). He one time wrote a paper, and I can't remember where it is. It was entitled: "Four or Five Different Ways to Publish a Mathematics Table." One way was to find fragments of the table at different places, and then carry out interpolations so as to get the increment and the independent variable the same for all these tables, round them off to the same precision, and then publish them without making reference to any of the previous authors. And this method, he said, is favored by the United States Coastal Geodetic Survey. (Laughter). Another way was to find fragments and then compute the likes. And that method is favored by people who he mentioned. But then, the least favorable technique was to find some defining expressions and do all the work yourself, and there were very few people like that. I was talking to Wilkes about Comrie's work, and he said that Comrie did things that nobody else had ever done in terms of computational techniques. You can find marvelous work like the Mathematical Tables" group, but they were using and they were grinding things out with Taylor series or something equivalent, and getting their precision by just a

6 Computer Oral History Collection, , lot of sweat and a lot of heavy labor. Comrie was just about the only one of his era who really tried to devise computational techniques in order to gain insight into what was happening with the tables he was constructing. You may know that he found an accountant sheet made by National Cash Register, and it had a number of registers in it that you could read from one register to another And he adapted that machine to table making and taught the young women to run it, after he left His Majesty's Almanac Office, and opened up Scientific Computing Services at #30 Bedford Square. Is that the posting machine that shifted at each register? That's right, that was the machine. There's another story about Comrie and I don't know whether it was the truth or not, it may be. At any rate, the Royal Navy needed some tables of some differential equations or something. During World War II, they made a contract with Cambridge, and also, to back it up, with Comrie. And the Math Department at Cambridge went to work on this, as did Comrie. Comrie worked on it for a few weeks and then turned in the results. So then the Royal Navy forgot they had another contractor working on it, and a year later, Cambridge walked in with a solution to the problem which they didn't need anymore, and the result being, that they invited Comrie to come to Cambridge and discuss the methods that he used. He took his machines with his girls and he went there and gave a lecture, and finished by saying, "Now this shows three things: First, these young ladies have never attended your university. And second, because they did not attend your university, they were able to learn some numerical methods, and third, because they didn't attend your university, they had not thought that what they knew was after all, quite indecent " (Laughter.) Of course, that must have been the attitude about computation at Harvard when you were first beginning to talk about machines. Oh, yes. The attitude was that a machine to do what I wanted to do would have so many parts that, based on elementary probability, some of the parts would after all, never work. That was one view.

7 Computer Oral History Collection, , Another view was that if you got it to work, it would only be a short time and you would have carried out, executed all the work that one could impose on such a device, and thereafter, it would be just a white elephant. It would be very much like a museum and a university knows that it should not breed too many museums. This led to Connant's letter to me, and I can't remember the date of it.. Connant wrote me a letter and said that I was one of these "faculty instructors". Do you remember that terrible time when people were supposed to instruct the faculty? (Laughter.) I was a "faculty instructor" with a three-year appointment. Connant's letter was to the effect that the committee that he had appointed to investigate had, and that I could look forward to being promoted in the near future and become an associate professor, but that I would never advance beyond that rank. That was a bitter letter. It was almost saying, "Get on to something else if " "Or get out". It's incredible, you know, the kind of manners people had. Leontief, when I interviewed him, had a letter when he came to Harvard, in which they said that they wanted him for certain things that he was doing, but as far as mathematical economics was concerned, they couldn't understand it, they didn't think it worth doing, and he wanted them to know that no one there would be interested in whatever results he might have obtained. That is the same kind of attitude, but I suppose, as we were talking you see, you see, before I went into physics, I was an undergraduate in mathematics, and I did also mathematics in my first year of graduate work, which took me through Now there was no one there who ever did any computing. The nearest that I was saying that anyone was interested in even numerical processes was G.D. Burkhoff, and he only because of his concern with the theory of finite differences. Did you ever run into that with him? Yes. But people like Marshall Stone, for example, and others, they never even calculated a definite integral.

8 Computer Oral History Collection, , Well, Professor Menzel told me that when he was interested in a computation in astronomy, Burkhoff was the only one on the faculty who talked to him about problems that he had. Well, there were a number of members of the Math Department even up to the time of my return, who had never been in the computer business. There was one who publicly stated that he had never set his eyes on one. (Laughter). Never had and never will. It's really quite amazing. A young lady, who has been doing some research into the revolution in which you were associated, came up with an interview with a Mrs. Rice, Mrs. Edmund Rice, who was apparently a student of yours at Radcliffe in those days of separation, in apparently 1935 to to 1936, that's when I was giving the first course in physics. They didn't do much in separate physics courses. I taught that course once too. Yes, they had the first course in physics, the lecture course demonstration. And Cathy always gave that course, and then he was away for a couple of years and I was given the opportunity of this. Then Harry Clark took it over. Yes.

9 Computer Oral History Collection, , And then, since I wasn't a real physicist, they decided that I could teach the girls. Well, this girl says, and I think she is crazy but I think we ought to get it from the horse's mouth, that you even then talked one day about the possibilities of what she called scientific calculation with a machine. She remembers you having a roll of paper under your arm, but I thought that was too early. I think so. Either she's wrong about the time, or I'm afraid this is just not right. Because you were then working on your thesis. Yes. This was the period in which I was beginning to form my future. I didn't know it of course, but this was the period in which I was beginning to form my future. But I can't imagine having conveyed those interests to the girl. It didn't sound right. Now, you know, there are a number of things in that period. There was this paper of Professor Menzel's on computing, but it has two dates on it: it has a 1935 date and a 1948 date, and you can't tell which was which, and you can't tell what his attitudes were towards computational needs, which I think is the essence of the paper. The need for better means of doing computation. I can tell you something about that. Harlow Shapley was, in one way or another, associated with IBM, and due largely to Ted Brown, I went to see Shapley. Shapley, as Shapley did in most anything. Shapley was a man who as long as what you said was not obviously nonsense, would give you a hearing and take interest, and if he could help, he would do it. So Shapley, in Harvard College Observatory, became interested in computing long before Menzel long before that. Because I can remember going out there to talk to Shapley and running into Menzel, and Menzel would be completely quiet as long as we were together and.

10 Computer Oral History Collection, , When I interviewed Ted Brown, he told me (I don't know whether I ought to call him Ted, but everybody does, so I slipped into it, you know. I don't want you to think I'm presuming to have more acquaintance with him than I have) that he was a student of the other Brown at Yale in mathematical astronomy. He went to a talk that Shapley gave after Shapley had seen the work of Eckert, not J. Presper Eckert, but Wallace, and talked about the possibilities of using machines, business machines and machines for computing, and that this interested him tremendously. But then when he came to Harvard, he originally wanted a job in mathematics and he saw G.D. Burkhoff and Burkhoff arranged for him to do part mathematics and part business school. Then, he told me about his acquaintance, and I guess he was one of Watson's favorites in many ways for a variety of reasons. But what I couldn't understand about his relation to you was, how did you get to Brown? How did I get to Brown? I could see that once you met him and he got you to talk to Shapley, and you'd come back and he'd get you to Watson, but it's hard to get your foot on your ladder. It's a good question and interesting. I recognized very early that the thing that would be required to make this computer go was money and a lot of it. By then, I had decided that I should build the first machine out of somebody's existing parts, rather than to start out with metal and wire or go ahead, but try and use somebody's pieces. Well, there were lots of different pieces to be used. There were the pieces of the computer industry at that time. There were the pieces of the step switches and so forth of the telephone industry. The telephone industry also had teletype and that was wire control printers, and they had punched paper tape, and I could foresee that it didn't make any difference whether you used tape or cards for input. The tape was harder to edit and you couldn't sort with it, but nevertheless, it would work and it had advantages. So all these different techniques printing telegraph techniques, telephone switching techniques, computer industry techniques were all griss for my mill, at that time, largely as a promoter, trying to find out where to get these pieces so that this machine could be put together. To this end, I went to the Monroe Calculating Machine Company. That was the first step, and what was the name of the charming man I met there? I'll tell you in just a minute.

11 Computer Oral History Collection, , Mr. Chase? Chase! My gosh, where did you get that? Let's see. Well, the only reason I happen to know about Chase is that Chase has published a history of mechanical computing machinery, and he describes how Aiken outlined to me the components of a machine which would solve his problems. His plans provided automatic computation and the four rules of arithmetic, pre-established sequence control, storage and memory of instold or computer values, sequence control which could automatically respond to computer results or symbols, together with a printed record of all that transpires within the machine, and a recording of all the computed results. I recognize the feasible construction in the outline of the mechanism he proposed to build. It was not restricted to any specific type of mechanism. It embraced a broad coordination of components which could be resolved by various constructive medium." He's just saying what I said a moment ago, only much better. Well, I just thought you might like to know where I got it. You said, how did I get it. Is this the paper? Yes, that's the one. That's the symposium. Okay. It's the symposium at Pittsburgh. It was John what's his name? He's dead now. John Bowman. Well, anyway, I went to Chase, and did just what he said. What's the date on that visit? Is that dated? Was that in 1937?

12 Computer Oral History Collection, , It was April 22, Okay, there it is. I went to him, and Chase was Chief Engineer at Monroe, and a very, very, scholarly gentleman. He took almost immediate interest, and we kept up an association for quite a few years there. He wanted in the worse way, to build Mark I. He would supply me with the parts and we would collaborate and do it together, that's what he wanted to do. He also foresaw what I did not. I did not foresee the application to accounting as coming out of it, and he did. He went to his management at Monroe and he did everything within his power to convince them that they should go ahead with this machine because, although it would be an expensive development, it would be invaluable in the company's business in later years. Now this is the thing that Tom Watson was never able to see. He was not. Oh, no. I'll mention that in just a few minutes. But Chase could see this. His management, however, after some months and months of discussion turned him down completely. They wouldn't have a thing to do with it. They said it was totally impractical. In fact, they sounded almost like the Harvard faculty. But in his moment of disappointment, it was he who said, "Go to IBM, because IBM is very successful and they more apt to undertake a thing of this kind than anyone else." And I said, "All right, who I should see at IBM?" And he said, "Why don't you see Professor Brown? He's at the Business School at Harvard. He's right there." So I said, "all right", and I went to see Ted Brown. Ted immediately came to life and suggested that I see Shapely. He immediately came to life and off to IBM. That's the way it happened. Now it's interesting that although IBM spent money for the parts and provided men to help with the reducing practice, none of the people who worked on this job or Watson could ever see that Mark I had any meaning whatsoever at all, other than an advertising venture to satisfy this peculiar guy from Harvard. The men who were detailed workmen found it interesting, but didn't want to spend much time on it. They felt that they had to work on something that had some commercial value, because if they didn't, on one of his trips to Endicott, Watson would find out that they hadn't produced anything in the way of a new product, and they'd be in

13 Computer Oral History Collection, , trouble. So that after the basic ideas had all been laid down and we'd gotten to the place of execution, I had to make a trip every month or so to Endicott to beg these people to get back on the job and do a little bit more. That's the reason it took seven years. Even after I went into the Navy, I still made those trips to Endicott to plead: "Come on, let's do a little more." That's very interesting. Again, you know, you are so apt to read things backwards, and they think of the involvement of IBM after their SSEC in the computer industry, and they just assume that this was their vision of it. But as I was saying to him, it seemed to me that I still don't fully understand why they did it, unless their parts were available. It would be, if it worked, a great feather in their cap. It would give them a lot of good publicity, they had a charitable instinct to some extent. Watson, despite all of his bad qualities, after all did have some idea of helping knowledge and new technology, and if this was a great feat that you could do, let's sort of do it. Was it more than that No. Absolutely not. Because one can see that none of those engineers, even the ones that IBM talks about in terms of their contribution Lake Hamilton and so forth And Durfee. And Durfee. You never hear of them doing anything in this line afterwards. That they just didn't decide that this was Well, I think you give IBM too much credit for vision. No, I'm not, you see.

14 Computer Oral History Collection, , No, but the vision didn't start until much later than the SSEC, it was in the 1950's. Oh, I see. That's their statement. It started after the defense calculator came into being. Oh, yes. IBM got going in the computer business when young Tom was made president. That's right. And as he told me one time, the first thing that he did when he became President of the Company, replacing his father, was to sense the embarrassment the Corporation faced because Remington Rand was getting all the credit for going ahead. So that if the senior Watson had remained active for say, another two years, chances are pretty good that Sperry Rand would be the big computer company today. It was changed by young Tom. That's right. He didn't have the technical knowledge so much as the vision. He had been exposed during the war, in aviation and to radar particularly, and I think this was where his knowledge of electronics came from. And he was also being needled by all the large customers: "What's Remington Rand doing? What are you going to do?" That sort of thing. So that Watson actually couldn't understand what all the publicity was about. He couldn't see that it had any practical value. He felt that he had been slighted when the credit was handed out, very much, he felt slighted. In fact, he felt so slighted, he tried to get me fired from Harvard, despite the fact that at no time had I ever tried to take any credit away from him.

15 Computer Oral History Collection, , Well, that's a terribly interesting thing to me, because that answers one of the always biggest questions I had of your getting to IBM and why IBM did it. By the way, since this is being recorded, I didn't mean to say that SSEC was a value judgment, but IBM always, in their official handouts, talk about going back to their SSEC, you know. I don't want to get into that, but that's another question and let's not worry about their later history. I was interested in another thing which was connected with that. Of course, Monroe would come to you. As I remember, we had in the top floor of Cruft, a whole battery of Monroe s, and had them at the Observatory. It was either Monroe or Marchant or sometimes Burroughs, but we all at Harvard, ran on Monroe s. It was an old one that I learned to compute, under Fred Whipple at the old laboratory which they knocked down, you remember, by the old tennis courts. They had old ones that were so old that the gears were worn. Instead of watching, you could listen and hear the thing turning over and count the clicks. (Laughter.) But in your thesis, you talk about the methods of solution of expansion in infinite series, and numerical integration by iterative methods. You obviously, whether it was experiencing your computing with a Monroe or what, immediately turned to the concept of digital computing. And that interests me because of two things and now let me tell you them, and you'll see how wrong I probably ought to be in my guesses. At that time, of course, the only people who were thinking in terms of computers were people who were doing analog computing, and while it's true that we've forgotten about that, after the war, Bush and his group did make a very complex analog computer taking over all the things that you had been developing, and that had come out of the ENIAC, and you can, after all, program an analog computer to do a great deal. And I've seen that as I look back, that if you look even as late as the forties, it wasn't clear to people as to whether the future lay with digital or analog. And during the war the development of fire control, radar, and so forth, bomb sites, all use analog computers and everybody is thinking about it. Well, it seems to me very odd that you should not, and especially so in that when I spoke to Vannevar Bush and asked him how he really got started on this, he said the big push came when they began to make power networks, and they were terribly worried about possible black-outs such as we eventually had. They couldn't do the computing, and they set up and began to devise these analog computers and that's how he said he got into the thinking about these differential analyzers and so on. Now, you had a back ground in electrical power engineering. You lived in a world of analog computers. It would seem to me to be the obvious conclusion of that simple syllogism that you should have been what you weren't. Well, as I said, I got going on non-linear differential equations, and non-linear differential equations are not the easiest to deal with on analog computers, and it was for that reason

16 Computer Oral History Collection, , that I began being interested in digital machinery. This was it, really, and this was the thing to do. Also, I guess there was another point. The analog machines are two or three digits, and I foresaw the enormous libraries of mathematical tables, not realizing that the very thing I wanted to create to make the Mathematical tables was necessary. (Laughter.) It's easier to compute your results than to look them up. Yes. And it was for this reason that the early phases of the computer had such a strong table connotation on the work that was done there. I remember as the youngest member of the Physics Department at lunch, your coming in with the first volume of the Bessel functions and tables, and everybody terribly excited by it. Well, going on with the astronomy, I suppose at a certain stage, you must have come on the work of Eckert. Perhaps not Comrie until later, but No, I found Eckert before Comrie. In fact, I went to see Eckert at Columbia. Somebody at IBM and I can't remember who Rice. And I can remember to this day walking into Eckert's office, leaving my overcoat on a chair in the anteroom, and going into the office and sitting down. After we'd been talking for four or five minutes, he said, "Where's your coat?" And I said, "It's on a chair on the second floor." And he jumped up and dashed up and retrieved my coat and brought it into the office, and he said, "You mustn't leave things lying around Columbia". (Laughter). Even then. Well, what was very clear to me when I saw Eckert's Laboratory and talked to him was that he had made excellent use of machine, even as Comrie did of the National Cash Register Machine. Eckert was not a man to redesign the machinery or

17 Computer Oral History Collection, , change the machinery. He was a man to use it and extract the maximum benefits from it. But what he was doing was computing. (Laughter.) Isn't that amusing to say it that way? Yes, except when you think about iterative procedures, you realize you can't go to vertical methods. Of course, there are so much more elegant means of expressing these ideas today, but it was to me. So I left Eckert's Laboratory with a feeling that as a competitor he was just doing fine. There was plenty for me to do. It was a problem that hadn't been solved. That's right. Did you have contact with Ben Wood in that same period? No, I never met him. He was a consultant to IBM and I had never met him, but I've heard about him. Because he had computational needs in handling all the data from all the tests that he was working with. There was a famous dinner that I've been told about. I was a graduate student, of course, and I knew nothing of it at the time, in 1938, and I was interested in the people who were there. The list seems to include Dean Westergard, of course, who was at the engineering school, E.B. Hunington, who I guess was still active then, and getting kind of old

18 Computer Oral History Collection, , Yes, this was a Harvard dinner. Yes, a Harvard dinner. There was Stern, and that must have been Ted Stern, who was, I suppose, the best of the young mathematical astronomers that they let go. This was the dinner in which I was asked to describe what I wanted to do. Right. There was Bridgeman. Yes. Funny choice, though I must say that I would love to tell Bridgeman anything and I always did. There was Ted Kimball, yourself, Brown, and J.G. Phillips. But what's interesting is that the only man from the Math Department there was E.B. Huntington, and I suppose it was because he had done statistical tables. He was more conscious of the fact that than anybody else in the Math Department. He was also more Yes. Actually, that dinner, which I had forgotten all about, was probably the. These were the kind of people.

19 Computer Oral History Collection, , No, as a matter of fact, I remember as you must certainly do, and I don't know if you are aware of this. After the War, at the time that you were holding those great conferences, and I don't remember whether this was the first one or the second one, but I remember that there was a great discussion between you and Norbert Wiener as to whether the computer was really a brain, and this kind of question which newspapermen like to ask, you know. Obviously, you took opposite sides on it. But I remember that I was lunching with some people in the Math Department and I don't remember who they were, but I remember that there was just general agreement that this had nothing to do with mathematics as the Mathematics Department could see it. They could never think of a mathematician being interested. I remember that somebody there and it might have been Garrett Burkhoff's eventual "von Neumann is", and somebody said, "yes, he was one of the greatest mathematicians of the twentieth century, but obviously, he was getting old." And I think that if there was one thing about which those mathematicians would have agreed, it was that. And if I'm not wrong, it was Marston Morse who said this, and he was a very able mathematician. Well, the way you describe the mathematician's attitude is accurate. I remember it even into the fifties. These things were for other people, but not for mathematicians. Well actually, this attitude prevails today. The computer establishments in the universities are almost never in the math department. Well, as I see it, the mathematicians weren't interested. The engineers were interested in the engineering problems of it, but they had yet not been taught to think in terms of what the computer might do. They solved their problems with big slide rules and tables and for them it was extraneous too. The only people that I can see as a group who really were interested were astronomers, and of course, as I was telling you, they were the big computers and the people who did the tables. Physicists didn't really do much computing except on a simple level. That's right. It was astronomers who essentially extended the techniques of Adams and Ashforth. Ashforth is a man whom I admired. Do you know about Ashforth?

20 Computer Oral History Collection, , No. You ought to look at his books sometime. There was a real brain and an unsung hero over the modern world. He was a great mathematical astronomer, who also designed machines, tide machines. Of course, they were all analog, but very interesting ones. And they solved the differential equations defining shaped drums. Yes. One question that I'd like to ask you about is a double one here. We've talked about the fact of digital, but my question then was: was it simply the IBM contact that drew you away from vacuum tubes? After all, you were Leon Chaffee's student, in a sense, and his life was a life of vacuum tubes. I take it there was some thought of quenching circuits of some kind that would use vacuum tubes in the Mark I, wasn't there at some stage? Yes, for radio radar parts. Well you question really is: having grown up in space charge in arbitrary-like Cruft, why was Mark I an electronic device? Again, money. It was going to take a lot of money. Thousands and thousands of parts, and using the techniques of the digital counters that had been made with vacuum tubes just a few years before I started, for counting cosmic rays, it was very clear that this thing could be done with electronic parts too. But what it comes down to is, if Monroe had decided to pay the bill, this thing would have been made out of mechanical parts. If RCA had been interested, it might have been electronic. And it was made out of tabulating machine parts because IBM was willing to pay the bill. Looking back, of course, if you can compare what you were doing with Mauchly and Eckert, they were getting speed and you were getting reliability. Isn't that what occurred to you then? Oh, yes. In fact,

21 Computer Oral History Collection, , Because you used to holler about reliability. The one word as I told him when he said, "What do you first remember about Aiken?" I said, "Reliability". Yes. You see, a computing machine is not reliable. It's worthless, absolutely worthless. You can't trust the results. And that was the reason that we checked everything that we did. Even today, people don't go to all lengths to check what they did. Just think of, nobody ever questioned it. Nobody ever questioned anything. I did an awful lot of checking even on the machines that we had, to be sure that there wasn't something. Well, we were conscious that we had and the ENIAC didn't. We had program facilities which they didn't. We had input and output which they didn't bother to worry about. And last of all, we had a precision so that if they used Simpson's Rule to integrate tiny little steps. And we could do the same integration with Mark I with twelve thicknesses, and then suddenly and make great big steps, you see, and we could do it once and they had to do several times to see if they could get the same number s twice. So that we used to say, "What's all this speed for? What does it accomplish? We get there sooner." And yet, you know, I really question if electronic computation would ever have become a great success had it not been for the transistor. I know. I've said it too, without being as wise as you are. There are people who argue against it because they point to Sage, because Sage did work with hand picked vacuum tubes, but it's the unusual one. It's that reliable transistor that really made the difference and made it go. It was that and the memory. I mean, the memory suddenly went from the least reliable to most reliable almost overnight. magnetic bubble memory just as sure as anything, and ten to twenty bits was not impossible. Just think of the problems that were going to get solved with that memory. You've heard all the talk about mechanizing the Library of Congress back in 1950, people were talking about that. Now let's mechanize the. All of this back in the fifties, there was consideration being given to problems of this kind.

22 Computer Oral History Collection, , There was research on information retrieval techniques. My student, Jerry Saltman, is one of the most distinguished people in this area. That began before we had the memory, and really these information retrieval techniques have been put together by Saltman and others over the last decade and will come into its own in its mechanization. I have no question a practicing attorney will have the law sitting on his desk with some keys and information retrieval techniques to assist him in making briefs. This is just going to happen so clearly. Accountants, lawyers, physicians you can imagine a physician having at his desk a key approach device that will give him all the side effects not of one drug but a combination of drugs. It will be invaluable. The Encyclopedia Britannica up there might be on your desk with some keys. Why will you want all those books? And as the new information comes and you revise it, you're up to date; you don't have to wait for the next edition. That's right, and you know that the chances are pretty good when the are reduced to practice, that you can build a device of that kind for less than it costs to make those books. I know you're not talking to me about what's coming in the future. Well, I really wanted to talk about the future as well as the past. Yes, well let me do my three more past, and I want to say that I think this is a fascinating thing, because I think it's astonishing always to see what your insights are. And it's not divorced from the past. No, it's not.

23 Computer Oral History Collection, , I had one question which I don't know really how much you want to put on record about, but it had always interested me about it, and that is what really I guess Bryce didn't contribute anything in the way of ideas. Bryce? He was a very astute inventor. A very astute inventor. So I gather. And he was a valuable senior advisor. If you started down a road that didn't look very practical, he could put his finger on it just about that quick. He was worth all the other IBM people put together. But Lake, Durfee and Hamilton, did they contribute any real ideas? Lake contributed absolutely nothing. He was the head of the department in which Durfee and Hamilton worked. Durfee contributed sweat. I must tell you an amusing story about these gentlemen, which I think will go far to make it clear to you. I went up to Endicott after this began to be formalized at IBM over the years. During the conversations with Lake and Durfee about what kind of machine this was that we were talking about. It was, oh I guess I made eight or ten such trips before I learned that IBM didn't know how to divide. And that was a terrible blow. They could add, they could multiply, but by god, they didn't know how to divide. It was almost like the bottom had dropped out. Maybe I ought to get back to Monroe they knew how to divide. So I stayed at a hotel in Binghamton and I can't remember the name of it. That night when I found out that they didn't know how to divide, I was up nearly all night, and it was that night that I invented the technique of dividing by computing by reciprocals. This is a scheme by which you can compute reciprocals, knowing how to add and multiply and you know how to do it. You could add and multiply by computing reciprocals and all you need is a first guess, and a first guess, when you are dealing with an independent variable that proceeds by fixed intervals, the first guess is always the reciprocal of the last name, and the convergence is beautiful. You double the number of digits of precision at each iteration.

24 Computer Oral History Collection, , So it was around three o'clock in the morning when this all came clear. The next day I walked into IBM and said "Well, you don't need to worry about not being able to divide because I know how to divide with an adder and a multiplier." And I started to derive this expression using the. And it became very clear that this was a waste of time because these men couldn't understand this. They knew no mathematics. Even high school algebra was too much for them. I came in the next day and said, "You know, it doesn't make any difference if you can divide because I can divide with an adder and a multiplier", and I started to derive an expression with this, which was meaningless. So then I said, "Well, you do this. You want to find the reciprocal for the number, and you guess any number whatever that you think is a first approximation, one digit of accuracy. You multiply these and you subtract from two, and then you multiply them a guess, and you'll have a number which is closer than the guess. Then you take that and repeat it all over again, and keep on going until you get the accuracy you wanted to get the reciprocal." Well, that helped, especially when the example I used was to compute the reciprocal of three on the assumption that three tenths was the first approximation. That would give you.33 and the next time,.3333 and up. Durfee and Hamilton thought this all over, and Durfee said that he wanted to go away and think about it. He disappeared and didn't show up until 5:00 that night. He came back with an enormous piece of drawing paper, and it was covered with pencil arithmetic on all sides, and he threw it down on the table in front of Hamilton, and he said, "Ham, it works every time." (Laughter.) I am so glad to know that's a true story. Harry Clark told me that. Do you remember old Harry Clark? I sure do. And he said to me, "And it works. It works every time." And then, on the basis that you can't patent mathematics this is what we began talking about Hamilton designed a mechanism to do this, and I believe he has a patent. He made an application and I suppose he has a patent on attaining the values of reciprocals for this thing. He used the credit, not me.

25 Computer Oral History Collection, , That took care of division. Yes. Well, now then there was a lot of talk at IBM about this, and Bryce, whom you mentioned, became confident that IBM ought to learn how to divide, and that being satisfied by approximating, he designed a dividing machine. This was then incorporated into Mark I. So we ended up having a Bryce divider in Mark I, which we took out three or four years later because we built a fast relay multiplier, a fast relay multiplier with this gimmick back here. And with that fast relay multiplier, we could divide faster than the dividing machine. This was the arithmetic unit right in here, and by taking out that divider, we had all that space and we could use the counter reels and increase the storage capacity, which went from 72 storage registers up to 100 and some. Oh boy, this was a tremendous. I had seen the piece of Mark I that's in the Comp Lab, but I hadn't seen that portion of it. The fast relay multiplier is not there. This is the relay multiplier, and this thing here, this cabinet had in it some subsidiary sequence devices that would run the machine for 22 instructions, the instructions being determined by plug board. And it also had in it a whirligig, an arm rotating in synchronism with the main shaft of the machine, so that a revolving lamp would (Interruption) Let's see, what were we talking about? We were talking about the rapid divider. Oh, yes. I was starting to say that we had this revolving lamp and we could put that across the cams that took the sparks and turned on the circuits of the relay and those cams would get burned. So Saturday morning was reserved for maintenance. One of the things that you did was to connect this thing and all the cams in sequence, and make sure that they were in pretty good shape so we could probably run for the next week.

26 Computer Oral History Collection, , I take it that although they did provide a divider other than the one you provided, that all of their contributions were of that kind. I mean, they were mechanical linkages and so on. Well, their contributions were arithmetic, using their techniques. It was pretty off the shelf, then, technology. Except for the one divider. Well, the divider became a standard technique in their technical machine design thereafter. Oh, it did? That was going to be my next question. And that's the Bryce. They never used your method. No. Did you ever write that up that you remember? I'm just interested, because methods of dividing happen to be a little hobby of mine. No, I don't believe I did. And it was never used so far as you know, either.

27 Computer Oral History Collection, , Oh, yes, it's been used quite a bit, but always by programmers. Quite a few programs have used it. You see, very few machines have square roots, and you use the same technique for square roots and reciprocal square roots. Yes, I'd say the technique has been very widely used, but not so much for division because most machines have dividing in them. I'm not surprised at your ingenuity in devising a method, but I was just a little curious too as a side, it isn't a major issue obviously, as to how you happen to think of using that rule? Had you had experience with it? I had had a good deal of experience with Newton graphs, because I had already recognized that I had to use Newton graphs in square roots. I see. Being interested in Newton, you see, that's what fascinated me. So having already decided that Newton graphs would be used for square roots, it wasn't too much of an extension to say, "Well, if I can't divide, well I'll see what I can do with Newton graphs." Well, that's extremely interesting and I think very valuable, because no one is really ever entirely sure about what contribution they made, and as you know, afterwards, that became such a silly question overshadowing the primary one. Well, one should not belittle what IBM did by any means. They had all this equipments that in totally different form was used, but it did do a performance of arithmetic, or Field's arithmetic, Bryce's then. And once those equipments were made available, then the overall organization of the machine and the design and techniques for printing, final printing, and the design of techniques for automatic control was about what was left that had to be done to put this thing together in one piece. So I wouldn't belittle what IBM did by any means.

28 Computer Oral History Collection, , Given the attitude at Harvard that I don't think is any secret even in terms of their attitude towards what you were doing, how did you convince Harvard to get involved in this venture with IBM? I suspect that by having gone so far that they couldn't quite see how to stop that. I've always thought that my experience at Harvard in that regard was very much like Hickman's experience. Hickman was a three-year instructor, and he was due to be told that this was the last appointment, when he found a Navy radar station transmitting and he could get this stuff service equipment. He went out and scrounged this equipment and presented it to Harvard, and Harvard didn't have the heart to say, "Gee, good-bye." So he managed to make a career out of it. And I think my experience with the University was about the same. Nobody wanted a computer and nobody thought it was going to be any good, but I had gone so far with the thing that it just didn't seem right to have me give it up. You know, it's really amazing. The war came and then once it was there, there was no more any problem, but I was present, again in my junior capacity, when it was decided to scrap the battery room to make room for the Mark I, which was due to come up. It was all ready to come on and there was no place to put it. And there was a long discussion: "Well, you know, do we really ", and "is this for?" I remember one question that was asked, I wouldn't remember by whom at that time, was, "Well, will there be money when we get rid of this and the wars over, to restore the battery room?" (Laughter) Well, we had a famous Harvard professor of physics, Joseph Lubbering, who built all kinds of things and was a great man in many ways. He was a professor who had the Bridgeman Chair, Nat Hollis' Chair, for fifty years, and at a meeting of the faculty, they asked whether they should introduce electrical engineering, and he said, "No". He said, "It's just a spurt. I've seen these things come and go."(laughter.) And that's the way people are, and that, I think you're not going to change. It may be that the Moore School was different, but I would really like to know what they thought down there. My guess is that there wouldn't be any academic group with a vision to see what this was. Well, who even outside of academic groups? In fact, I was going to ask you how many people other than yourself really did have a vision as to what the many, many kinds of things computers could do other than just tables? It was pretty slim. Comrie could see it. He saw it from one end to the other.